Title: Change-point analysis in high-dimensional econometric dynamic factor models
Authors: Ansgar Steland - RWTH Aachen (Germany) [presenting]
Abstract: A new approach to test for a change in the covariance structure is proposed especially targeting high-dimensional econometric factor models. The approach is based on bilinear or quadratic forms of CUSUM statistics combined with a multiple testing procedure. Contrary to existing tests that fail and/or are computationally infeasible when it comes to high dimensions $p$, the proposed methodology can even be used for $p > T$. The class of factor models allowed for covers many specifications used in econometric data analysis and modeling. It even allows for an infinite number of correlated factors. The approach is carefully examined by simulations. We illustrate the approach by analyzing the impact of the Covid crash on the Fama-French factors.